CFA: Correlation > 1 for latent varia... PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 William Welch posted on Wednesday, December 09, 2009 - 8:02 pm
Dear Mplus -

I ran this first model and I received no error warnings:

ANALYSIS: estimator=ml;

MODEL:
Commit by OC1 OC2 OC3;
Beliefs by Company Company2 Company3;
NewBond by Newcon1 Newcon2 Newcon3 Newcon4;
NewSA by NewSAE1 NewSAE2 NewSAI1 NewSAI2 NewSAC1 NewSAC2;
FVal by FVal1 FVal2;
FSoc by FSoc1 FSoc2 FSoc3 FSoc4;
FEnhanc by FEnhanc1 FEnhanc2 FEnhanc3;
FProtec by FProtec1 FProtec2 FProtec3;
FUnders by FUnders1 FUnders2 FUnders3;
FCareer by FCareer1 FCareer2 FCareer3;

OUTPUT: standardized mod(3.84) tech4;

However, when I ran a second, reduced model, I received a "correlation greater than 1 for latent variables" warning:

ANALYSIS: estimator=ml;

MODEL:
Commit by OC1 OC2 OC3;
NewSA by NewSAE1 NewSAE2 NewSAI1 NewSAI2 NewSAC1 NewSAC2;
FVal by FVal1 FVal2;
FSoc by FSoc1 FSoc2 FSoc3 FSoc4;
FEnhanc by FEnhanc1 FEnhanc2 FEnhanc3;
FProtec by FProtec1 FProtec2 FProtec3;
FUnders by FUnders1 FUnders2 FUnders3;
FCareer by FCareer1 FCareer2 FCareer3;

OUTPUT: standardized mod(3.84) tech4;

I checked the latent variable correlation matrix in the tech 4 output, and there are latent variables that have a correlation great than 1.

Is this a solvable problem, or is this model not a viable one?

-Will
 Linda K. Muthen posted on Wednesday, December 09, 2009 - 11:24 pm
Factors that correlate one or greater are not statistically distinguishable and the model should be modified.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: